Comparison of Offline Monitoring Algorithms for Production Data of Workshop Equipment Based on DTW

Current data acquisition and analysis algorithms based on electrical signals are generally solved using distance similarity algorithms, for which a Dynamic Time Warping (DTW) based algorithm is proposed. The similarity distance generally needs to calculate the distance between sampling points and points, and then carry out subsequent calculations. For the problem that the similarity distance of the existing basic electrical signal cannot determine the threshold, a dynamic time-based regularization algorithm is given. Firstly, the similarity distance was calculated by using Euclidean and Fréchet algorithms for different time periods after preprocessing, and the similarity threshold was found to be 86. Then the dynamic time regularization algorithm is introduced to obtain the similarity distance threshold of 40. Finally, aligning the time domain and current amplitude domain of the algorithm can improve the similarity distance accuracy of anomaly identification under offline monitoring. The experimental results show that the Euclidean distance is susceptible to the influence of extreme data in a more complex workshop environment, and the Fréchet algorithm will have a large error, which is suitable for a more standard and uniform model of each component, while the proposed improved algorithm can accurately calibrate the threshold value from 40 to 35.